Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Tue, Jul 22, 2008 at 03:38:41PM -0600, Gregor Mosheh wrote:
> Gregor wrote:
>>> A few fewer errors, but still not even able to run a verify. What'd
>>> be the next thing?
>
> Hey, I got it! This came to me when I saw a posting about writes
> arriving on the DRBD target out of sequence and causing trouble. They
> were talking about ReiserFS and we use ext3, but ... we're using 'tc'
> the Linux traffic shaper. We use SFQ in addition to HTB, and SFQ's
> mission is to randomize ("perturb") packet queues to keep any one
> session from hogging traffic.
>
> My hypothesis: Do not use DRBD in conjunction with tc, at least not
> while using SFQ, or else it'll reshape your sync traffic, screwing it up.
>
> Now that I have turned off the traffic shaping (we're shopping for an
> appliance to do a better job than tc anyway) the verify is running
> smoothly. I have even re-enabled CSO and TSO on the network adapters,
> and haven't seen a burp yet.
interessting empirical observation.
but I really doubt any causality there.
if SFQ (or any other tc thingy) would cause tcp transmission errors
or data alteration, that would be a so serious bug someone
else should have noticed by now.
--
: Lars Ellenberg http://www.linbit.com :
: DRBD/HA support and consulting sales at linbit.com :
: LINBIT Information Technologies GmbH Tel +43-1-8178292-0 :
: Vivenotgasse 48, A-1120 Vienna/Europe Fax +43-1-8178292-82 :
__
please don't Cc me, but send to list -- I'm subscribed